Repetition blindness refers to the finding that report of both words from a sequence presented at rates of 8-10 per second is reduced if one of the words is a repetition of an earlier one. Reduced report also occurs even when the pair of words is merely similar orthographically or phonologically. We have found that if the items are pronounceable nonwords or pseudohomophones rather than words, repetition improves recall. This result was not obtained using a variant of the task. We report experiments which contrasted tasks, methods and types of items and varied lexicality and the relationship between the critical item pairs. The critical pair of items was repeated, similar or unrelated words, nonwords, homophones or pseudohomophones. The pattern of effects differs for different tasks and when the critical items are merely similar rather than identical. We consider the relationship between repetition and similarity effects and explanations of repetition blindness.